Systematic review screening
An include or exclude decision, with written reasoning, for every paper — for |
The evidence
Abstractus finds more of the relevant literature than a single human reviewer does, and it does it while you are still opening the first file.
Relevant papers found
Abstractus
Manual
screenings Abstractus completes in the time a researcher finishes one
Time to screen a review
Measured on public benchmark datasets, alongside published data from “Single-reviewer abstract screening missed 13 percent of relevant studies: a crowd-based, randomized controlled trial”, Gartlehner et al., 2020. Our methodology is written up on the benchmark results post.
What you get
Every abstract gets an include or exclude decision with the reasoning written out, and every number a PRISMA diagram asks for comes out with it.
Upload your search results and set your inclusion and exclusion criteria. Every record comes back with an include or exclude decision, however many thousands of them there are.
Each call carries the written reasoning behind it, so a decision can be checked, argued with, or overruled. Nothing is a black box, and nothing rests on trusting the tool.
Duplicates removed, records screened, decisions and their justifications: the numbers a PRISMA flow diagram asks for, exportable to CSV whenever a reviewer asks for them.
Search
A review is only as good as the search behind it. Abstractus builds the search strategy, then runs it across the literature and the patent record in one pass.
Free. Turn a research question into ready-to-run search strings for the databases you search yourself, with verified MeSH and Emtree terms and a live hit count so you know how broad a strategy is before you commit to it.
Or skip the copying and pasting. One question goes to every database at once and comes back as a single deduplicated set, with the exact query we sent to each one shown beside its results.
We search these for you
Research Databases
· 6 databases
Searched at once, merged and deduplicated, abstracts ready to screen.
Patents and trademarks
· 3 registers
US patents from the same full-text index USPTO's own examiners search, and international patents from the EPO's worldwide collection of roughly 100 authorities.
We write the search string, you run it
Every major review database
· 14 databases
Each one in its own syntax — controlled vocabulary, field tags, proximity operators and truncation rules all differ, and getting any of them wrong silently changes what the search returns.
How it works
The same workflow you already follow, with the reading compressed from months into minutes.
Your question, your inclusion and exclusion criteria. Search Assist turns that into search strings for the databases that suit it.
RIS, NBIB or CSV from any database. Metadata is normalised and duplicates are removed before anything is screened.
An include or exclude decision for each paper, with the reasoning written out, so a second reviewer can check the call, not just the outcome.
Sort, filter and read the reasoning behind any paper. You stay the reviewer of record; the tool removes the volume, not the judgement.
Decisions, reasoning and PRISMA counts out to CSV, ready for your methods section and your co-authors.
From the blog
Benchmarks, methodology, and notes on evidence synthesis. Our numbers are written up so you can check them.
13 August 2026
Screening is the biggest part of a review, but it is not the only part. Recent updates: Search Assist, deduplication, and collaboration next.
Read the post
1 August 2026
On a public benchmark of 57,177 abstracts, our screening reaches a median recall of 95% and cuts the manual screening workload by 84%.
Read the post
25 July 2026
We are starting a blog about systematic reviews, evidence synthesis, and the software we are building to make screening faster.
Read the post
Bring your search results and screen them today, or talk to us about seats for a department or research group.